forkprobe: An AI Skill Selection and Trial Run Tool for DeepSeek Harness
jayden-x-l/forkprobe
ForkProbe is a DSH plugin that runs multiple AI skills side-by-side on the same task, generates a local report, and lets you pick the winner to continue.
Run multiple candidate skills side by side in DeepSeek Harness, generate a local HTML report, and let you see real outputs before choosing a winner.
dsh plugin --profile web add github:jayden-x-l/forkprobe- Category
- Skills & Agents
- Platform
- DSH-Plugin
- Author
- jayden-x-l
- Distribution
- Plugin
forkprobe Key Features
forkprobe Repository
forkprobe Repo Summary
ForkProbe is a DeepSeek Harness (DSH) plugin for AI skill selection and trial runs, maintained by Jayden-X-L under the MIT license, last updated in 2026-08. It solves the problem that when there are more and more skills online, you no longer worry about "whether there is a skill" but "which skill should be used for the current task"—instead of trusting skill descriptions, it runs the same task with the model itself and multiple candidate skills side by side, generates a local HTML report, and lets you see real outputs before choosing a winner. ForkProbe supports scenarios like academic polishing, naturalization rewriting, reviewer response, and PPTX generation, and recommends candidates from locally installed skills, EverMind Skill Hub, GitHub, or BYO paths.
What core features does the ForkProbe DSH plugin support?
ForkProbe's core features include: running baseline and multiple candidate skills in parallel, showing each run's full output, time, token estimate, file preview, and AI review suggestions; after you select a winner, the Report's "Continue" button saves a local handoff and lets the Agent continue the task along the winning skill; it also supports anonymously sharing the skill selection to accumulate samples for future community recommendations. v0.10 adds a native DeepSeek Harness plugin forkprobe-dsh, which uses the native forkprobe_compare tool to launch candidate subagents in parallel, without nesting a second dsh process or copying DSH credentials.
How to install ForkProbe?
As a DSH plugin, ForkProbe can be installed with the following command:
dsh plugin --profile web add github:jayden-x-l/forkprobe
Before installation, user confirmation is required, and local skills are scanned and candidates deduplicated for safety.
How to use ForkProbe?
You don't need to remember commands. Just tell the Agent: "Compare a few skills first and see which one fits the current task better." Or more explicitly: "Use forkprobe to recommend candidates, wait for my confirmation, then run them side by side and generate a report, and let me choose the winner." The English trigger is: "Compare a few skills first and see which one fits the current task better."
What scenarios is ForkProbe suitable for?
ForkProbe is suitable when you are unsure which skill to use for the current task and want to see real outputs before deciding, especially for deliverables like PPTX, research figure packages, survey reports, runnable web pages, or video productions. It is not suitable for simple deterministic tasks—if the answer or tool path is already clear, direct execution is faster.
Which DeepSeek Harness versions and platforms does ForkProbe support?
ForkProbe is a native DeepSeek Harness plugin, depending on official packages like @deepseek-ai/dsh-tools, @deepseek-ai/dsh-agent, @deepseek-ai/dsh-llm, and @deepseek-ai/dsh-subagent, deeply integrated with the DSH ecosystem. The project is maintained by Jayden-X-L under the MIT license, last updated in 2026-08, with current version v0.10.